ACS Chem Neurosci. 2017 Apr 19;8(4):785-797. doi: 10.1021/acschemneuro.6b00371.

In Vivo Systems Response Profiling and Multivariate Classification of CNS Active Compounds: A Structured Tool for CNS Drug Discovery.

Waters S1,2, Svensson P2, Kullingsjö J2, Pontén H1, Andreasson T3, Sunesson Y3, Ljung E2, Sonesson C2, Waters N2.

1 Department of Pharmacology, Gothenburg University , SE-405 30 Gothenburg, Sweden.
2 Integrative Research Laboratories Sweden AB , Gothenburg SE-413 46, Sweden.
3 AstraZeneca R&D , SE-431 83 Mölndal, Sweden.

Abstract

This paper describes the application of in vivo systems response profiling in CNS drug discovery by a process referred to as the Integrative Screening Process. The biological response profile, treated as an array, is used as major outcome for selection of candidate drugs. Dose-response data, including ex vivo brain monoaminergic biomarkers and behavioral descriptors, are systematically collected and analyzed by principal component analysis (PCA) and partial least-squares (PLS) regression, yielding multivariate characterization across compounds. The approach is exemplified by assessing a new class of CNS active compounds, the dopidines, compared to other monoamine modulating compounds including antipsychotics, antidepressants, and procognitive agents. Dopidines display a distinct phenotypic profile which has prompted extensive further preclinical and clinical investigations. In summary, in vivo profiles of CNS compounds are mapped, based on dose response studies in the rat. Applying a systematic and standardized work-flow, a database of in vivo systems response profiles is compiled, enabling comparisons and classification. This creates a framework for translational mapping, a crucial component in CNS drug discovery.

KEYWORDS:

Neurochemistry; antipsychotics; biomarkers; dopidines; drug discovery; monoamines; multivariate data analysis; phenotypic screening; response profiles; rodent behavior; systems pharmacology; translational modeling

PMID: 27997108

 

Supplement

ISP represents a novel approach to CNS drug discovery, with the potential to significantly improve the discovery process by reducing time and resource needs as well as development risks.

The technology emanates from an integrative view on disease states and drug induced states, ie drug effects. Based on this systems level perspective, the pharmacological evaluation is focused primarily on comprehensive, phenotypic characterization in vivo. The application of a structured methodology for such phenotypic characterization is described in the paper. We argue that the approach presented provides not only a major step towards more efficient CNS drug discovery, but also a means for gaining better understanding of CNS drug actions as such.

Traditional drug discovery processes are essentially built around the aim to create drug candidates that bind strongly and selectively to specific molecular targets, preferentially single targets, such as receptors, transport proteins, or enzymes. Such binding is then assumed to lead to therapeutic benefit. Accordingly, there is a strong focus on assessment of binding to various targets, typically performed in artificial in vitro preparations disassembled from the full physiological environment in which the molecular target is expressed in the disease state of interest. The outcome in terms of novel CNS treatments using this strategy has not met expectations. A number of reasons for this are discussed in the paper.

Essentially, for complex physiological systems, such as the CNS, there is typically no clear-cut, simple correspondence between modulation of a single target protein in vitro, and the functional outcome at the system level, ie in the real life situation. This can be due to network mechanisms, such as feed-back or feed-forward mechanisms making the system robust to disturbances at isolated targets or network nodes.  Also, the target protein as such may not respond in the same way to the drug in a test tube, as it does in the full physiological environment. Furthermore, the significance of different aspects of the in vitro binding as such, eg affinity, kinetics, or down-stream signaling, for different targets is far from clear.  For instance, for dopamine D2 receptor ligands, extensively studied for decades, the functional impact of association/dissociation kinetics, receptor dimerization, and biased signaling remains to be elucidated.  In general, the knowledge on detailed mechanisms leading to CNS disorders is incomplete, as is the knowledge on therapeutic mechanisms, which means that it is not possible to reliably predict real-life drug effects based on reductionistic theories and test-tube measurements. This applies not only to therapeutic effects, but also to adverse effects. Hence the limited success for target-based drug discovery strategies in the CNS area may be an inherent problem of the strategy as such.

On the other hand, it has been shown that when an alternate approach, referred to as phenotypic screening, ie studies on an integrative level, in living systems, is applied, success rates in drug discovery and development programmes are higher.

ISP, Integrative Screening Process, was developed as an effort to evolve the concept of phenotypic screening, closely inter-linked with drug design, to a new level of rational, systems oriented drug discovery.  The underlying principle is to systematically assess drug effects in integrated, physiological systems, which makes the findings translatable to human conditions to a higher degree. We collect large amounts of data, both on neurochemical and gene related biomarkers, and also on functional outcomes, as reflected by behavioral patterns. The data-array collected for each test compound constitutes a broad “response profile”. We take great effort to standardize and optimize experimental conditions, and to collect data for a wide range of CNS active compounds, to map known property space.

The current ISP database covers all major CNS therapeutic classes, and all major classes in terms of the presumed molecular mechanism of action, ie traditional receptor binding profiles, in addition to novel test compounds.  By means of advanced data mining and machine learning techniques, we can link the response profiles both to chemical properties of the molecules, and to clinical effects as observed in human. Hence, we obtain information on the type of profile we want to achieve for novel compounds, as well as strong handles on how to design the compounds chemically.  Thus, based on our response profiling we have created a research engine that can efficiently generate novel candidate drugs with desired effects, as well as provide improved insights regarding how drugs in clinical use affect the CNS in real life.

To date, nine candidate drugs have been generated using the ISP technology.

Ultimately, this type of systems response approach will likely be necessary to substantially improve our understanding of how CNS drugs work, and how they impact CNS disease states, on the molecular target level as well as on an integrated systems level.  There are also ethical aspects, in so far as a systems pharmacology approach maximizes the utility of each experiment performed.  Further, the systems profiling strongly favors druggable compounds, which means that less resources, including animal studies are spent on compounds with druggability issues.

The paper discusses theoretical considerations, as well as the practical application of ISP, and provides examples of compounds discovered and characterized by the principles outlined.

 

Selected reading

Lee, J. A. and Berg, E. L. (2013) Neoclassic drug discovery: the case for lead generation using phenotypic and functional approaches J. Biomol. Screening 18, 1143– 1155

Maggiora, G. M. (2011) The reductionist paradox: are the laws of chemistry and physics sufficient for the discovery of new drugs? J. Comput.-Aided Mol. Des. 25, 699– 708

Sams-Dodd, F. (2005) Target-based drug discovery: is something wrong? Drug Discovery Today 10, 139– 147

Swinney, D. C. and Anthony, J. (2011) How were new medicines discovered? Nat. Rev. Drug Discovery 10, 507– 519

Tun, K., Menghini, M., D’Andrea, L., Dhar, P., Tanaka, H., and Giuliani, A. (2011) Why so few drug targets: a mathematical explanation? Curr. Comput.-Aided Drug Des. 7, 206– 213

Sykes, D. A., Moore, H., Stott, L., Holliday, N., Javitch, J. A., Lane, J. R., and Charlton, S. J. (2017) Extrapyramidal side effects of antipsychotics are linked to their association kinetics at dopamine D2 receptors, Nat Commun 8, 763.